[PLUGIN] bzr changeset now support rollup changesets (and bzr send-changeset)

[Martin Pool]

On 29 Jun 2005, Aaron Bentley <aaron.bentley at utoronto.ca> wrote:
> Martin Pool wrote:
> > * The base is redundant with the first parent and can be removed. 
> 
> This introduces the constraint that the changeset can never be relative
> to anything except the first parent.  For this branch
> A -> B -> C -> D

Sorry, my mistake.  Both seem to be needed.  We can, at least, avoid
sending the SHA1 of the revision twice because it must be the same.
Perhaps we can avoid identifying the specific base if there's only one
parent.

> > * File ids and parent ids shouldn't normally need to be transmitted: we
> >   know precisely which revision this should be applied to, therefore
> >   precisely which filename has what id.  File-ids only need to be
> >   given in the changeset for newly added files or directories.
> 
> Yes, I think you're right about file-ids.  What about text-ids?  I've
> never been clear on their status.  Are they intended to be local-only
> data?  If not, then since they're part of the inventory SHA-1, they must
> be transmitted unless they can be computed, right?

In the current code you'd need the text-id to generate the right
inventory, but they don't really add any value and are a bit of an
accident of the current implementation.  From talking to Robert is seems
like a more useful thing to put in there is the revision-id where the
text was introduced (i.e. the last one where it was changed.)

-- 
Martin